On-line Planar Area Segmentation from Sequence of Monocular Monochrome Images for Visual Navigation of Autonomous Robot

We introduce an on-line segmentation of a planar area from a sequence of images for visual navigation of a robot. We assume that the robot moves autonomously in a man-made environment without any stored map in the memory or any markers in the environment. Since the robot moves in a man-made environment, we can assume that the robot workspace is a collection of spatial plane segments. The robot is needed to separate a ground plane from an image and/or images captured by imaging system mounted on the robot. The ground plane defines a collision-free space for navigation. We develop a strategy for computing the navigation direction using a hierarchical expression of plane segments in the workspace. The robot is required to extract a spatial hierarchy of plane segments from images. We propose an algorithm for plane segmentation using an optical flow field captured by an uncalibrated moving camera.